Joint hashing yields repeated exact owner outputs
Provedmme_released_joint_interior_hashed_owner_output_restrictmatrix-multiplicationtensor-extraction
Joint parent hashing followed by physical owner regrouping restricts the released owner-window product to exact owner outputs with precisely the product of the six post-repair copy counts. The exact output identities of the hashing steps are explicit hypotheses supplied by the existing step construction. The root exponent bound remains a separate obligation.
Preamble
import Theorems.Thm_mme_bigAdd_mono_restrict import Theorems.Thm_mme_released_joint_interior_hashed_product_restrict import Theorems.Thm_mme_released_joint_interior_owner_output_product_restrict open scoped BigOperators open MME MME.TensorObj MME.RecursiveYZ MME.CompleteSplit open MME.ReleasedJointInterior MME.MoreAsymmetryExactSeed universe u
Formal statement
theorem mme_released_joint_interior_hashed_owner_output_restrict
{K : Type u} [Field K] (k : ℕ) (hk : 0 < k) (eps : ℝ)
(E : ∀ r : Fin 6, ProfiledCW.ExactStep 2 (blocks r k * 4) (gradedSource r k eps))
(a : ∀ r : Fin 6, Address 4 270 (parent r) (size r k))
(houtput : ∀ r, (E r).output = fun i x =>
Graded (parent_total r) i (a r)
(ProfiledCW.split (positions r k) (positions_length r k) x) ∧
Useful (fullCell (parent_total r) (a r)) (integerProfile r k i)
(ProfiledCW.split (positions r k) (positions_length r k) x))
(e : ∀ j : Fin 270, Fin ((k * weight j * denominator ^ 4) * 2) ≃
Position (fun r => size r k j))
(length : ∀ j : Fin 270,
((k * weight j * denominator ^ 4) * 2) * 2 ^ (2 - 1) =
(k * weight j * denominator ^ 4) * 4) :
let P : ∀ j : Fin 270,
ProfiledCW.Predicate ((k * weight j * denominator ^ 4) * 4) :=
fun j i z => 0 < weight j →
(∀ p : Fin (k * weight j * denominator ^ 4),
(∑ q, (ProfiledCW.split (ell := 3) (Equiv.refl _) rfl z p q).val) =
ReleasedInterior.parent (component j).2 0 ((roleEquiv (component j).1).symm i)) ∧
∀ w : CompleteWord 3,
|(Fintype.card {p : Fin (k * weight j * denominator ^ 4) //
ProfiledCW.split (ell := 3) (Equiv.refl _) rfl z p = w} : ℝ) /
(k * weight j * denominator ^ 4 : ℕ) -
((((ReleasedGlobal.jointRows (component j).1 (component j).2).map
(fun p => if ReleasedGlobal.atom p.1
((roleEquiv (component j).1).symm i) = w then p.2 else 0)).sum : ℕ) : ℝ) /
(denominator : ℝ) ^ 4| ≤ eps
let O : ∀ j : Fin 270,
ProfiledCW.Predicate ((k * weight j * denominator ^ 4) * 4) :=
fun j i x =>
Graded (ReleasedInterior.parent_total (component j).2)
((roleEquiv (component j).1).symm i)
(fun r t => (splitEquiv r j).symm (a r j t))
(ProfiledCW.split (e j) (length j) x) ∧
Useful (fullCell (ReleasedInterior.parent_total (component j).2)
(fun r t => (splitEquiv r j).symm (a r j t)))
(fun c w => k * weight j * ReleasedInterior.integerProfile
(component j).1 (component j).2 ((roleEquiv (component j).1).symm i) c w)
(ProfiledCW.split (e j) (length j) x)
Restrict
(bigAdd (fun _ : Fin (∏ r, (E r).copies) =>
kronFin 270 (fun j => ProfiledCW.tensor K (O j))))
(kronFin 270 (fun j => ProfiledCW.tensor K (P j))) := by sorrySource
Checked tensor restrictions and released global histogram extraction.